It is typical for modern VLSI circuits to have a very large number of outputs and, in particular, a very large number of outputs of scan paths. DE 195 36 226 A1 discloses how errors can be identified by supplying an external comparison vector in the test mode, by direct comparison of the actual test results at the outputs of the circuit with a comparison vector which is supplied to the circuit. When an error is identified, the result of the bit-by-bit comparison is emitted, once again bit-by-bit, in order to diagnose the result. In this case, the respective externally supplied comparison vector has the same word length as the test response vector which is emitted from the tested circuit and as the test results which are produced at the circuit output. The procedure described in the aforementioned document is accordingly complex for circuits having a large number of outputs. The long word length of the supplied test vector which is required in this case, the bit-by-bit comparison for the large number of bits to be compared and, in the case of diagnosis, the outputting of the large number of bits are in this case complex. This becomes clear if one considers circuits with about 1,000 to 2,000 outputs which, for example, may be outputs of scan paths. These outputs must be monitored during testing and, all the more, during diagnosis. Furthermore, the number of test input vectors is frequently on the order of several hundred to several thousand test vectors. A further disadvantage is that this method cannot be used in all situations.
DE 101 37 745 describes the storage of diagnosis data on a chip. However, this is not feasible with modern chips, owing to the additional area requirement.